Simulations were used in two main parts: first, as a proof of concept before any experimentation had taken place to determine what experiments might provide meaningful information; and second, to assist in answering questions arising during analysis of experimental results. As part of the benchmarking analysis, the co-heating tests were simulated on the building models described in Section 4.4.5. This uses the co-heating test as a calibration tool for the models and confirms that observed HTC is included in the array of models. The building models were set up to use observed internal temperature, and the weather file created from the Bureau of Meteorology, to simulate the co-heating tests as conducted in field on each configuration. The simulation results were then analysed using the multiple regression analysis described in Section 5.3.3 so they can be directly compared with field results. This approach accounts for the solar gains and the wind conditions observed during the test period.
Simulation Results
Simulation results are grouped by configuration to show how changes to the thermal shell have influenced the HTC estimate. The uncertainty for the simulations has been taken as the 95% confidence interval of the daily variation in HTC, an approach used by Stamp (2011).
This has been selected over an analysis based on altering data inputs as simulations do not have the same uncertainty due to the measurement instruments. The uncertainty presented in this fashion more closely represents the uncertainty inherent in the co-heating analysis, rather than the entire co-heating experiments.
157 5.6.1.1 Configuration 1
Figure 5.25 Simulated HTC estimates, Configuration 1
Figure 5.25 shows several simulations match closely with the observed co-heating test HTC, with a good spread above and below. There is a clear difference between the models with and without increased ceiling insulation, but no variation between the models with and without increased thermal mass.
Base Model 4 is qualitatively the best representation of the test cell based on the material data and the infiltration rate entered. This model overestimates the HTC by 24.5%. The best estimate of the HTC from the Base set is Model 12 however, which has R2.5 insulation in the walls instead of the estimated R1.91, and an infiltration rate of 0.52 ACH. The nearest matching estimate overall is from Ceiling Model 8 (R-Value 1.5, infiltration 0.866 ACH). This indicates the building performs better than expected based on initial estimates of thermal shell characteristics. This supports the observed difference, reported in Section 5.5, between co-heating test HTC and the heat flux test HTC.
158 5.6.1.2 Configuration 2
Figure 5.26 Simulated HTC estimates, Configuration 2
As in Configuration 1, Figure 5.26 shows many of the models have HTCs similar to that estimated for the test cell. The best estimate from the base model set is Model 11, which has wall insulation of R1.5 and an infiltration rate of 0.52 ACH. The nearest estimate is from Thermal Base Model 11, though the difference between this and the base model is only 0.1 W/K. The best qualitative representation of the test cell, Base Model 4, overestimates the HTC by 15.3%.
159 5.6.1.3 Configuration 3
Figure 5.27 Simulated HTC estimates, Configuration 3
Figure 5.27 demonstrates the same patterns within Configuration 3 as were observed for Configurations 1 and 2. The best estimate from the base model set is Model 10 (R-value 1.91, infiltration 0.52 ACH). The best estimate overall is Thermal Base Model 11 (R-value 1.5, infiltration 0.52 ACH). The best qualitative representation of the test cell, Base Model 4, once again overestimates the HTC by 15.4%.
Discussion of Simulated Results
All three configurations show large differences between the sets with and without ceiling insulation, but almost no difference between sets with and without increased thermal mass.
The steady-state nature of the co-heating test is designed to minimise the influence of thermal mass by eliminating the charge and discharge of energy throughout the day. These buildings are expected to show differences in dynamic behaviour, but the steady-state behaviour ultimately assesses the overall U-value of the building, regardless of whether it is a heavy or lightweight construction.
160 All four model sets (Base, Ceiling, Thermal Base and Thermal Ceiling) gave at least two models inside the uncertainty range from the field co-heating tests. These demonstrate how varied the characteristics of a building can be, and yet still have the same HTC, as well as showing that the simulated dataset covers a good range of building performances compared with the field testing.
The initial building model, Base Model 4, in each configuration overestimates the heat loss by more than 15%. This is in line with observations reported in Section 5.5 that the heat loss determined by heat flux and blower door testing was much higher than heat loss determined by the co-heating test. As information from this analysis informed the initial model, it is not surprising that heat loss is overestimated. Reducing theoretical losses by increasing R-values or reducing infiltration rates improves alignment between the simulated co-heating test and the field co-heating test. This indicates there is some part, or parts, of the building fabric that is being underestimated.
The purpose of the simulation analysis with respect to the overall research aim is to show how the decay method will identify the performance of different buildings. These simulations provide a combination of models that have the same HTC for different thermal shells, and different HTCs. This provides a valuable resource for analysing how the decay method will identify the buildings with different HTCs as well as what differences are observed when two differing models have the same HTC. The observation that the thermal mass makes little difference to the HTC is particularly important as it is expected that this will change the building’s dynamic behaviour, providing good comparison between the analysis of the decay method and the co-heating test.
161